mox/smtp/data.go
Mechiel Lukkien 1f9b640d9a
add faq for smtp smuggling, fix bug around handling "\nX\n" for any X, reject bare carriage returns and possibly smtp-smuggling attempts
mox was already strict in its "\r\n.\r\n" handling for end-of-message in an
smtp transaction.

due to a mostly unrelated bug, sequences of "\nX\n", including "\n.\n" were
rejected with a "local processing error".

the sequence "\r\n.\n" dropped the dot, not necessarily a big problem, this is
unlikely to happen in a legimate transaction and the behaviour not
unreasonable.

we take this opportunity to reject all bare \r.  we detect all slightly
incorrect combinations of "\r\n.\r\n" with an error mentioning smtp smuggling,
in part to appease the tools checking for it.

smtp errors are 500 "bad syntax", and mention smtp smuggling.
2024-01-01 20:11:16 +01:00

181 lines
4.8 KiB
Go

package smtp
import (
"bufio"
"bytes"
"errors"
"io"
)
var ErrCRLF = errors.New("invalid bare carriage return or newline")
var errMissingCRLF = errors.New("missing crlf at end of message")
// DataWrite reads data (a mail message) from r, and writes it to smtp
// connection w with dot stuffing, as required by the SMTP data command.
//
// Messages with bare carriage returns or bare newlines result in an error.
func DataWrite(w io.Writer, r io.Reader) error {
// ../rfc/5321:2003
var prevlast, last byte = '\r', '\n' // Start on a new line, so we insert a dot if the first byte is a dot.
// todo: at least for smtp submission we should probably set a max line length, eg 1000 octects including crlf. ../rfc/5321:3512
buf := make([]byte, 8*1024)
for {
nr, err := r.Read(buf)
if nr > 0 {
// Process buf by writing a line at a time, and checking if the next character
// after the line starts with a dot. Insert an extra dot if so.
p := buf[:nr]
for len(p) > 0 {
if p[0] == '.' && prevlast == '\r' && last == '\n' {
if _, err := w.Write([]byte{'.'}); err != nil {
return err
}
}
// Look for the next newline, or end of buffer.
n := 0
firstcr := -1
for n < len(p) {
c := p[n]
if c == '\n' {
if firstcr < 0 {
// Bare newline.
return ErrCRLF
} else if firstcr != n-1 {
// Bare carriage return.
return ErrCRLF
}
n++
break
} else if c == '\r' && firstcr < 0 {
firstcr = n
}
n++
}
if _, err := w.Write(p[:n]); err != nil {
return err
}
// Keep track of the last two bytes we've written.
if n == 1 {
prevlast, last = last, p[0]
} else {
prevlast, last = p[n-2], p[n-1]
}
p = p[n:]
}
}
if err == io.EOF {
break
} else if err != nil {
return err
}
}
if prevlast != '\r' || last != '\n' {
return errMissingCRLF
}
if _, err := w.Write(dotcrlf); err != nil {
return err
}
return nil
}
var dotcrlf = []byte(".\r\n")
// DataReader is an io.Reader that reads data from an SMTP DATA command, doing dot
// unstuffing and returning io.EOF when a bare dot is received. Use NewDataReader.
//
// Bare carriage returns, and the sequences "[^\r]\n." and "\n.\n" result in an
// error.
type DataReader struct {
// ../rfc/5321:2003
r *bufio.Reader
plast, last byte
buf []byte // From previous read.
err error // Read error, for after r.buf is exhausted.
// When we see invalid combinations of CR and LF, we keep reading, and report an
// error at the final "\r\n.\r\n". We cannot just stop reading and return an error,
// the SMTP protocol would become out of sync.
badcrlf bool
}
// NewDataReader returns an initialized DataReader.
func NewDataReader(r *bufio.Reader) *DataReader {
return &DataReader{
r: r,
// Set up initial state to accept a message that is only "." and CRLF.
plast: '\r',
last: '\n',
}
}
// Read implements io.Reader.
func (r *DataReader) Read(p []byte) (int, error) {
wrote := 0
for len(p) > 0 {
// Read until newline as long as it fits in the buffer.
if len(r.buf) == 0 {
if r.err != nil {
break
}
// todo: set a max length, eg 1000 octets including crlf excluding potential leading dot. ../rfc/5321:3512
r.buf, r.err = r.r.ReadSlice('\n')
if r.err == bufio.ErrBufferFull {
r.err = nil
} else if r.err == io.EOF {
// Mark EOF as bad for now. If we see the ending dotcrlf below, err becomes regular
// io.EOF again.
r.err = io.ErrUnexpectedEOF
}
}
if len(r.buf) > 0 {
// Reject bare \r.
for i, c := range r.buf {
if c == '\r' && (i == len(r.buf) || r.buf[i+1] != '\n') {
r.badcrlf = true
}
}
// We require crlf. A bare LF is not a line ending for the end of the SMTP
// transaction. ../rfc/5321:2032
// Bare newlines are accepted as message data, unless around a bare dot. The SMTP
// server adds missing carriage returns. We don't reject bare newlines outright,
// real-world messages like that occur.
if r.plast == '\r' && r.last == '\n' {
if bytes.Equal(r.buf, dotcrlf) {
r.buf = nil
r.err = io.EOF
if r.badcrlf {
r.err = ErrCRLF
}
break
} else if r.buf[0] == '.' {
// Reject "\r\n.\n".
if len(r.buf) >= 2 && r.buf[1] == '\n' {
r.badcrlf = true
}
r.buf = r.buf[1:]
}
} else if r.last == '\n' && (bytes.HasPrefix(r.buf, []byte(".\n")) || bytes.HasPrefix(r.buf, []byte(".\r\n"))) {
// Reject "[^\r]\n.\n" and "[^\r]\n.\r\n"
r.badcrlf = true
}
n := len(r.buf)
if n > len(p) {
n = len(p)
}
copy(p, r.buf[:n])
if n == 1 {
r.plast, r.last = r.last, r.buf[0]
} else if n > 1 {
r.plast, r.last = r.buf[n-2], r.buf[n-1]
}
p = p[n:]
r.buf = r.buf[n:]
wrote += n
}
}
return wrote, r.err
}